Distinct patterns of gene and protein expression elicited by ...

BMC Genomics

BioMed Central

Open Access

Research article

Distinct patterns of gene and protein expression elicited by organophosphorus pesticides in Caenorhabditis elegans John A Lewis*1, Maria Szilagyi2, Elizabeth Gehman3, William E Dennis1 and David A Jackson1 Address: 1US Army Center for Environmental Health Research, Fort Detrick, MD, USA, 2US Environmental Protection Agency, Washington, DC, USA and 3Battelle National Biodefense Institute, Frederick, MD, USA Email: John A Lewis* - [email protected]; Maria Szilagyi - [email protected]; Elizabeth Gehman - [email protected]; William E Dennis - [email protected]; David A Jackson - [email protected] * Corresponding author

Published: 29 April 2009 BMC Genomics 2009, 10:202

doi:10.1186/1471-2164-10-202

Received: 7 October 2008 Accepted: 29 April 2009

This article is available from: http://www.biomedcentral.com/1471-2164/10/202 © 2009 Lewis et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract Background: The wide use of organophosphorus (OP) pesticides makes them an important public health concern. Persistent effects of exposure and the mechanism of neuronal degeneration are continuing issues in OP toxicology. To elucidate early steps in the mechanisms of OP toxicity, we studied alterations in global gene and protein expression in Caenorhabditis elegans exposed to OPs using microarrays and mass spectrometry. We tested two structurally distinct OPs (dichlorvos and fenamiphos) and employed a mechanistically different third neurotoxicant, mefloquine, as an outgroup for analysis. Treatment levels used concentrations of chemical sufficient to prevent the development of 10%, 50% or 90% of mid-vulval L4 larvae into early gravid adults (EGA) at 24 h after exposure in a defined, bacteria-free medium. Results: After 8 h of exposure, the expression of 87 genes responded specifically to OP treatment. The abundance of 34 proteins also changed in OP-exposed worms. Many of the genes and proteins affected by the OPs are expressed in neuronal and muscle tissues and are involved in lipid metabolism, cell adhesion, apoptosis/cell death, and detoxification. Twenty-two genes were differentially affected by the two OPs; a large proportion of these genes encode cytochrome P450s, UDP-glucuronosyl/UDP-glucosyltransferases, or P-glycoproteins. The abundance of transcripts and the proteins they encode were well correlated. Conclusion: Exposure to OPs elicits a pattern of changes in gene expression in exposed worms distinct from that of the unrelated neurotoxicant, mefloquine. The functional roles and the tissue location of the genes and proteins whose expression is modulated in response to exposure is consistent with the known effects of OPs, including damage to muscle due to persistent hypercontraction, neuronal cell death, and phase I and phase II detoxification. Further, the two different OPs evoked distinguishable changes in gene expression; about half the differences are in genes involved in detoxification, likely reflecting differences in the chemical structure of the two OPs. Changes in the expression of a number of sequences of unknown function were also discovered, and these molecules could provide insight into novel mechanisms of OP toxicity or adaptation in future studies.

Page 1 of 21 (page number not for citation purposes)

BMC Genomics 2009, 10:202

Background The wide use of organophosphorus (OP) based pesticides and unresolved issues in their toxicity, including the causes of persistent and off-target effects and the mechanisms of neuronal degeneration, make them an important concern for public health. OPs are a class of chemicals that inhibit serine esterases by covalently bonding with the active site serine. Two primary targets of OPs have been implicated in human toxicity, acetylcholinesterase (AChE; reviewed in [1]) and neuropathy target esterase (NTE; reviewed in [2]). However, the inhibition of AChE is of more concern because of acetylcholine's role as a neural transmitter. Long-term adverse effects of OP exposure have been described [3-5], but the nature and mechanism of persistent effects are relatively poorly understood. The principal risk of toxicity from OPs and other AChE inhibitors occurs after high level, acute exposures when death from respiratory failure may rapidly ensue; less severe exposures may cause salivation, lacrimation, incontinence, and convulsions followed by paralysis potentially resulting in death (reviewed in [6,7]). However, a number of persistent and delayed effects of OP exposure are also known. A so-called intermediate syndrome–defined by weakness of the neck, proximal limb, and respiratory musculature–may present 24–96 hours after exposure and is believed to be the result of acetylcholine receptor desensitization (reviewed in [1,8]). Organophosphate induced delayed polyneuropathy (OPIDP) is a delayed syndrome (7–21 days after exposure) that is characterized by numbness, weakness, and paresthesia in the limbs and degeneration of peripheral nerves and central nervous system myelin sheaths; inhibition of NTE is thought to underlie OPIDP (reviewed in [1,8,9]). Chronic neurological and neuropsychiatric effects–some of which may persist for years–and developmental neuro-behavioral effects have also been described [10-12]. In an effort to understand the mechanisms of OP toxicity, we have tracked global gene and protein expression after intoxication by two OPs, dichlorvos and fenamiphos, using the genomic model organism Caenorhabditis elegans with whole genome microarrays and mass spectrometrybased proteomics. We selected two chemically different OPs to ask whether it is possible to distinguish between the biological responses to different inhibitors of AChE. To discriminate generalized alterations in gene expression due to neurotoxicity and stress from OP specific effects, we included a third chemical, mefloquine, as an outgroup. Mefloquine is believed to cause neurotoxicity by perturbing Ca++ homeostasis, most likely through interference with an ion channel [13,14]. Using C. elegans for toxicological studies provides a number of benefits. The organism is well studied, has a

http://www.biomedcentral.com/1471-2164/10/202

very simple body plan, and has a completely sequenced genome. Further, the responses of C. elegans to a number of toxicants have been shown to resemble those of mammals in a number of cases ranging from anesthetics to metals to OP pesticides [15-21] (see also [22] for a recent review of the uses of C. elegans in toxicological research), and the availability of commercial microarrays has facilitated the investigation of the mechanism of action of an array of toxicants at the functional genomic level (e.g., [23-25]). C. elegans does not require neuronal signals for respiration and is very resistant to death via OP intoxication yet shows substantial similarity to mammals in the relevant biochemistry and genomics [20]. The acute toxicity of OP pesticides results from inhibition of AChE in vertebrates [26] and in nematodes [20]. The C. elegans genome also contains two homologs of the vertebrate secondary OP target, NTE (ZK370.4 and M110.7; [27] and unpublished observations). While it is unknown whether inhibition of either of the C. elegans NTE homologs will induce an OPIDP-like condition, the syndrome has been described in humans following dichlorvos exposure (reviewed in [28]) raising the possibility that dichlorvos might be a suitable compound for investigating this effect. Furthermore, because C. elegans is resistant to OP lethality, we reasoned that by using this organism to study the effects of dichlorvos and fenamiphos, it might be possible to expose the worms to high doses of OPs to highlight changes in gene and protein expression that are difficult to discern using classical methods or animal models that are less resistant to OPs. A drawback to using C. elegans, however, is that the worms are usually cultured with bacteria as food source [29]. The presence of bacteria may complicate the interpretation of data because of the metabolism of test materials by the feeder organisms and the contamination of protein and nucleic acid samples with bacterial molecules. While a number of axenic media have been previously described (for example [30-36]), nematodes cultured in axenic media have generally shown reduced rates of development and extended life-spans, suggesting that the media lack essential nutrients. To overcome this problem, we developed a defined, liquid, sterile medium (CeHR medium) [37] in which C. elegans can be stably propagated with a generation time similar to that of worms on bacterial plates [37,38]. In this study, we exposed developmentally synchronized C. elegans cultures in CeHR medium to two structurally different OPs, dichlorvos and fenamiphos, and the functionally dissimilar neurotoxicant, mefloquine, as an outgroup. Global gene expression was determined by microarray analysis of RNA from harvested worms, and proteins




extracted from parallel worm cultures were analyzed by mass spectrometry to identify changes in protein expression. Proteomic and functional genomic analysis revealed sets of genes and proteins that distinguish not only between exposure to the OPs and to mefloquine, but also between the OPs themselves. The results are generally consistent across the transcriptomic and proteomic analyses and can readily be understood in the context of the known effects of OP intoxication.

Methods Nematode culture C. elegans [N2 wild type, DR subclone of CB original (Tcl pattern I), obtained from Caenorhabditis Genetics Center] were maintained in synchronized cultures grown in CeHR medium (see below). All cultures were grown at 22.5°C with shaking at 70 rpm on an Innova 2000 platform shaker (New Brunswick Scientific, Edison, NJ). Typically, 5 × 105 L1 larvae were used to inoculate 40 mL of medium in a T-75 flask. Stock cultures were propagated using the synchronization procedure described below to ensure that sufficient numbers of developmentally synchronized worms were available for experimentation at all times. CeHR medium is a sterile, defined medium, supplemented with 20% (v/v) ultrapasteurized organic, fatfree milk for the axenic propagation of C. elegans. A detailed description of the preparation of the medium is available from the USACEHR on request and in [37]. Synchronization of cultures Embryos were isolated using a minor modification of the bleaching method of Stiernagle [39] described by Szilagyi et al. [37]. The isolated embryos were suspended in 30 mL M9 buffer (42.3 mM Na2HPO4, 22.0 mM KH2PO4, 85.6 mM NaCl, 1 mM MgSO4), transferred into T-75 culture flasks and incubated at 22.5°C overnight to allow hatching and arrest at the L1 stage. L1 larvae were used within three days to start developmentally synchronized cultures.

Rangefinding A developmental inhibition assay was used to determine exposure concentrations. Synchronized worms grown at 22.5°C with shaking at 70 rpm progress from the midvulval L4 larval stage to the early gravid adult (EGA) stage within 24 h. The presence of toxicants inhibits this development. To determine concentrations corresponding to effect concentrations (EC) of EC10, EC50, and EC90, (concentrations preventing 10%, 50%, and 90% of the worms from developing to EGA), 8 × 104 L1 larvae were inoculated into T-25 flasks–each containing 10 mL of CeHR medium. When 90% of the worms had developed to midvulval L4 larvae (44–46 h), chemical was added. The flasks were incubated for 24 h, after which a sample of worms was examined microscopically to assess their developmental stage. The toxicant concentrations corresponding to EC10, EC50, and EC90 were selected for the exposure experiments (Table 1). Exposures L1 larvae (2.5 × 105) were suspended in T-75 flasks containing 30 mL CeHR medium and grown until 90% of the population had developed to the mid-vulval L4 larval stage–two flasks were allotted for each condition to provide adequate biomass for RNA and protein preparation. The worms were treated with mefloquine (Ash Stevens, Inc., Detroit, MI), dichlorvos, or fenamiphos (Chem Service, Inc., West Chester, PA) for 8 h or allowed to develop as a control; a sample was taken for chemical analysis to verify exposure concentration (Table 1). Each exposure was repeated three times.

Worms were harvested by centrifugation (800 × g for 3 min at 4°C), and the supernatant was aspirated. Samples for protein extraction were washed once with 0.1 M NaCl, centrifuged (800 × g for 3 min at 4°C), and the supernatant was aspirated. The pellets for protein and RNA extraction were suspended in the residual liquid, flash frozen by

Table 1: Concentrations of test chemicals

Chemical

Developmental Arrest (%)

Nominal Conc. (mg/L)

Average Conc. (mg/L)

Rep 1 Conc. (mg/L) Rep 2 Conc. (mg/L) Rep 3 Conc. (mg/L)

dichlorvos

10 50 90

3 15 50

3.55 16.0 52.7

3.65 16.5 53.4

3.44 15.9 53.0

3.57 15.8 51.8

fenamiphos

10 50 90

10 60 200

6.33 29.2 74.4

7.65 28.1 86.2

5.97 25.6 68.7

5.37 33.8 68.3

mefloquine

10 50 90

10 250 500

10.3 240 492

8.5 205 530

11.3 248 477

11 267 470



drop-wise addition to liquid nitrogen, and stored at 80°C. Chemical analysis Chemicals were analyzed on a Hewlett-Packard Model 6890 Gas Chromatograph equipped with a 6890 model series auto injector. Ions were measured for fenamiphos with a 5973 Mass Selective Detector, for mefloquine with a flame ionization detector, and for dichlorvos with an electron capture detector. Analytical standards were purchased from Chem Service, Inc. RNA methods Extraction and labeling Frozen worm droplets were pulverized in liquid N2 using a pre-chilled mortar and pestle. The pulverized worms were transferred to 6 mL Trizol (Invitrogen, Carlsbad, CA) and homogenized in a dounce homogenizer. RNA was purified according to the manufacturer's protocol and precipitated with isopropyl alcohol. After centrifugation, the RNA pellet was dried, dissolved in water, and subjected to an additional round of purification using the RNeasy Maxi Kit (Qiagen, Valencia, CA) according to the manufacturer's directions. The quality and yield of the preparation was assessed throughout processing and labeling using a 2100 Bioanalyzer (Agilent, Santa Clara, CA), and when necessary, the mass yield was confirmed using an ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington, DE).

Poly(A)+ RNA was isolated from the total RNA using OligoTex (Qiagen) essentially as described by the manufacturer. Two micrograms of poly(A)+ RNA (adjusted for rRNA contamination) was used as the template for cDNA synthesis using the SuperScript Choice Kit (Invitrogen) per the manufacturer's recommendations except that (1) a high pressure liquid chromatography (HPLC)-purified T24T7 promoter primer (Integrated DNA Technologies, Coralville, IA) was used to initiate first strand synthesis; (2) the second strand synthesis was not terminated using EDTA since we found that EDTA carryover interfered with subsequent enzymatic manipulations; and (3) PelletPaint (Novagen, Madison, WI) was used in place of glycogen for precipitation. Biotin labeled cRNA was synthesized from the T7 promoter incorporated in the cDNA using the BioArray High Yield RNA Transcript Labeling Kit (Enzo Life Sciences, Farmingdale, NY) per the manufacturer's recommendations; approximately 1 μg of cDNA was used for synthesis. cRNA was purified from unincorporated nucleotides and other reaction components using the RNeasy Mini Kit (Qiagen). Microarrays cRNA samples were hybridized to C. elegans whole genome GeneChips (Affymetrix, Santa Clara, CA), proc-


essed, and scanned at the Walter Reed Army Institute of Research Vaccine Genomics Laboratory, Rockville, MD using Affymetrix instrumentation and with hybridization, washing, and scanning parameters provided by the manufacturer [40]. Microarray data analysis Microarray data was processed using the robust multiarray averaging method (RMA) [41]. To verify inter-replicate reproducibility, replicate samples were subjected to pairwise correlation analysis of all probe sets. For the vast majority of replicate pairs, the R2 value was greater than or equal to 0.95, and no replicates were included with R2 < 0.92. A Present, Absent, or Marginal call for each probe set was determined using the R statistical package [42] and the Bioconductor [43] implementation of the Affymetrix MAS 5.0 algorithm (affy package 1.12.2). Only probe sets with at least three present calls in the complete data set were retained for further analysis. This procedure removed 5,623 out of the total 22,624 probe sets on the microarray from the analysis. We have observed that even when a multiple test correction is used in ANOVA with high dimensional microarray data, small differences in gene expression that are not credible on careful inspection of the signal intensities can be assigned highly significant p values. To reduce the impact of this problem, we retained a final tally of 4,999 probe sets that passed the Present/ Absent screen and changed by at least 1.8 fold from control for statistical analyses. Support vector machine for dosing standardization On inspection, the standardized concentrations of dichlorvos seemed to exert relatively greater effects on the patterns of gene expression in exposed worms than mefloquine or fenamiphos, yielding a right shifted pattern of gene expression (see Figure 1). To confirm this observation, we used a support vector machine (SVM, [44] Partek Pro Genomics Suite 6.0–default settings) to predict an apparent concentration (control, low, mid, or high) for each chemical to which the worms had been exposed based on patterns of gene expression. For SVM modeling, we used data from worms exposed to cadmium and acrylamide in parallel experiments (unpublished data) in addition to fenamiphos and mefloquine. No dichlorvos data were included. To take the differences between the measured and targeted concentrations of chemicals (Table 1 and not shown) into account for this analysis, we calculated an adjusted measure of developmental arrest by prorating the target level of arrest (10%, 50% or 90%) by the ratio of the measured concentration of toxicant to the target concentration (Equation 1). The 100 probe sets with the highest partial correlations to this adjusted value were used to train the SVM.




probe sets differentially affected by OP exposure, but unaffected by mefloquine.

3&

&KHPLFDO

'LFKORUYRV )HQDPLSKRV

3UHGLFWHG /RZ &RQF &OXVWHU

0HIORTXLQH 3UHGLFWHG +LJK &RQF &OXVWHU

&RQFHQWUDWLRQ &RQWURO /RZ 0HGLXP

3UHGLFWHG &RQWURO &OXVWHU

+LJK

3UHGLFWHG 0LG &RQF &OXVWHU

3&

Figure 1 of samples Clustering concentrations of dichlorvos, of worms fenamiphos, exposed to and standardized mefloquine Clustering of samples of worms exposed to standardized concentrations of dichlorvos, fenamiphos, and mefloquine. Principal components analysis plot depicting the clustering of samples of worms exposed to standardized concentrations of the three toxicants based on gene expression levels. Three replicates are shown. Nominal concentration classes are indicated in color, and chemical exposure groups (including respective controls) are indicated by shapes. Low, medium, and high concentrations refer to EC10, EC50, and EC90 from the developmental inhibition assay. A support vector machine (SVM) trained on a data set lacking dichlorvos exposed worms was used to classify the samples based on gene expression levels. Samples which the SVM predicts to have the same concentration level are joined by lines to a centroid. The percent variance in the data explained by each principal component is shown in parentheses.

adjusted arrest level = target arrest level ×

measured concentration target concentration

(1) Identification of OP specific gene changes For identification of OP-specific gene changes, we removed the fenamiphos low concentration and the dichlorvos high concentration data from consideration. The samples for the two remaining exposed concentrations for each OP were grouped based on the SVM classification as either "mid concentration OP" or "high concentration OP." The OP control samples and all of the mefloquine samples were placed into a third "no OP" class. An ANOVA identified 500 probe sets that are significantly different (FDR ≤ 10-4; false discovery rate, [45]) among the three classes. To eliminate those genes whose expression was even marginally affected by mefloquine exposure, we next removed probe sets that changed by 1.5 fold or more from control at any concentration, in any replicate of the mefloquine data, to generate a list of 94

Following statistical identification of the 94 differentially expressed probe sets, we inspected their mapping on the C. elegans genome (based on WormBase oligo mapping; WormBase release 180) [46] and found 20 of them that represent genes with at least one additional probe set on the microarray that was not identified, based on our strict criteria, as a probe set specifically affected by OP exposure. In most cases, these probe sets have similar patterns of expression but display slight differences in the magnitude of the fold change from control with the result that one probe set passed the fold change or statistical cut off while the other(s) did not. In two instances, the selected and rejected probe sets targeted different splice variants of the same gene. In another, the probe set showed a response to mefloquine, but the change was in the opposite direction compared to the OP responses; we deemed this to be an OP specific gene change. In a final case, the unidentified probe set had a signal intensity below background (indicated by no Present calls). We retained all 15 of these probe sets. However, we excluded probe sets for five genes each recognized by two probe sets. For four of these genes, one but not the other of the probe sets showed changes in expression upon mefloquine exposure with no readily apparent explanation for the differences. The other one hybridized to two genes, and we could not resolve which gene was being measured. One final probe set was removed because it was called Present (by MAS 5.0 algorithm) in only two of the OP exposed samples; the third sample in which it was called Present was a mefloquine sample. After these adjustments, a group of 88 probe sets (representing 87 genes) that respond to OP but not mefloquine exposure remained (Table 2, Figure 2). Gene ontology analysis In order to assist in interpreting the microarray data, DAVID [47-49] and GoMiner [50,51] were used to assess whether particular gene ontology terms occurred more frequently than expected by chance in the set of genes specifically affected by OP exposure. DAVID was run using the high stringency setting and the following annotation groups: Molecular Function level 4–5, Cellular Component level 4–5, Biological Function level 4–5, InterPro terms, and PIR keywords. Of the 88 probe sets submitted all but 9 were clustered by annotation. GoMiner was run through the web interface with default settings except that all ontology terms were used. The group of 88 probe sets that are specifically affected by OP exposure was compared to the annotation of the entire C. elegans genome for both DAVID and GoMiner for statistical evaluation. For



DAVID analysis, we report the negative antilog of the Group Enrichment Score as p values. Identification of OP discriminating gene changes For identification of gene changes that discriminate between the OP exposures, only data from OP exposed samples and their respective controls were included. As above, SVM predicted concentrations were used for classification and the low fenamiphos and highest dichlorvos exposures were omitted. A 2-way ANOVA using the SVM predicted concentrations and exposure chemical (fenamiphos, dichlorvos, and control) revealed 28 probe sets with significantly different expression between the two OPs (FDR ≤ 10-4). Two of these genes are also targeted by additional probe sets which do not meet the fold difference criterion but are similar in expression pattern to the originally identified probe sets, so both original probe sets were retained. This list was further refined to include only probe sets which changed by at least 1.8 fold as a result of the exposure and between chemicals resulting in a final list of 24 probe sets, representing 23 genes.

Microarray data have been deposited in the Gene Expression Omnibus [52], Accession Number GSE12298. Protein Methods Complete details of sample processing, mass spectrometry, and data analysis may be found in Additional File 1: ProteinMethods.pdf. A brief description follows. Purification and processing Frozen worm droplets from the highest concentrations of fenamiphos and dichlorvos exposures and the unexposed controls were ground in liquid N2 and resuspended [40 mM Tris, 1 mM EGTA, and 1 × Protease Inhibitor Cocktail (Sigma-Aldrich, St. Louis, MO)]. The suspension was sonicated, clarified by centrifugation, and lyophilized. Four milligrams of protein from each sample were denatured in 8 M UREA and dithiothreitol and then acetylated with iodoacetamide. After dilution to 1 M urea, the samples were digested with trypsin (Promega, Madison, WI). Peptide analysis The digested peptides were desalted, dried under vacuum, reconstituted in 10% acetonitrile, and fractionated using mixed mode ion chromatography with a Polycat A column and Polywax LP column in series (PolyLC Inc., Columbia, MD). Eight time based fractions were collected. Each fraction was analyzed using a nanoACQUITY UPLC coupled to a QTOF Premier quadrupole, orthogonal acceleration time-of-flight tandem mass spectrometer (Waters, Milford, MA). Data were collected over the 50– 1990 mass to charge (m/z) range using the Waters Protein Expression MSE method, which alternates between low energy scans to survey the precursor ions and high colli-


'LFKORUYRV 0LG +LJK

)HQDPLSKRV 0LG +LJK

WWU GKS )( WWU )* KVS PDS )( ))